AutoConfig: Autonomous TSN Configuration via Traffic Profiling

Abstract

Time-Sensitive Networking (TSN) enables deterministic communication in converged industrial networks. Configuring TSN end stations remains challenging due to dynamic traffic patterns and the lack of explicit application-level requirements. Many legacy and modern applications cannot communicate their quality-of-service (QoS) needs, limiting the effectiveness of existing TSN configuration workflows. This paper presents AutoConfig, an automated end-station configuration framework that infers traffic characteristics directly from observed packet streams and dynamically applies TSN scheduling policies. AutoConfig operates without requiring application modifications and introduces an offline/online separation, where traffic profiles are learned through monitoring and subsequently used for fast runtime adaptation via profile matching and switching. Experimental results show that AutoConfig achieves up to 100% detection accuracy for stationary traffic patterns and 83% for dynamic scenarios, while maintaining low configuration latency (around 12 ms) and modest system overhead on a Linux-based end station.